VORTRAG Jens Hocke:

Representation Learning: From Feature Weighting to Invariance

An ever increasing amount of data is collected. In many domains, it has become so large that an automatic analysis is mandatory. I present approaches to improve classification results by changing the representation of the input data. New linear representation learning methods, as well as new architectures for learning invariant representations, are discussed. 


Ort: Seminarraum des INB